Abstract

In spite of their significance in freight modeling, freeway design and operation, varying truck flow patterns by season and time-of-day cannot be captured by current truck data sources such as surveys or point detectors. In this paper, a truck tracking algorithm was developed to estimate path flows of trucks by a linear data fusion method utilizing weigh-in-motion and inductive loop point detectors. The authors utilized a Selective Weighted Bayesian Model (SWBM) that tracks individual vehicles between two detector locations using truck physical attributes and waveform signatures. Selected truck features were identified and weighted via Bayesian modeling to improve vehicle matching performance. Data for model development were collected from two WIM sites in California, separated by 27 miles. The algorithm showed a high matching accuracy for the truck population tracking across longer distance. In a test data set, the model was able to successfully match 76 percent of trucks that traversed the corridor. Although only 21 percent of trucks observed at the downstream site traversed the corridor, only 18 percent of the matches predicted by the model were false matches. In a follow-up case study, the algorithm was implemented over a longer 65-mile distance of freeway section and showed that the proposed algorithm was capable of providing insights into truck travel patterns and industrial affiliation to yield a comprehensive truck activity data source.